Interoperability is accepted as a fundamental necessity for the successful realization of Healthcare Information Systems. It can be achieved by utilizing consistent standards defining syntactic and semantic meaning of the information being exchanged. HL7 FHIR is one of such open standards for Health Information Exchange (HIE). While HL7 FHIR supports Representational State Transfer (REST) architecture and Service-oriented Architecture (SOA) for seamless information exchange, it inherits the inflexibility and complexity associated with the RESTful approach. GraphQL is a query language developed by Facebook that provides promising techniques to overcome these issues. In this paper, we exploit the use of GraphQL and HL7 FHIR for HIE; present an algorithm to map HL7 FHIR resources to a GraphQL schema, and created a prototype implementation of the approach and compare it with a RESTful approach. Our experimental results indicate that the combination of GraphQL and HL7 FHIR-based web APIs for HIE is performant, cost-effective, scalable and flexible to meet web and mobile clients requirements.

1. A GraphQL approach to Healthcare Information
Exchange with HL7 FHIR
Suresh Kumar Mukhiya, Fazle Rabbi, Ka I Violet Pun,
Adrian Rutle, Yngve Lamo
� skmu@hvl.no
November 5, 2019
(HVL) Health Information Exchange November 5, 2019 1 / 34

2. Who are we?
Our aim is to improve public mental health with Adaptive
technologies (ICT) and psychological treatments.
Team of Domain experts, Software Engineers, ML engineers, HCI
engineers and Industry.
(HVL) Health Information Exchange November 5, 2019 2 / 34

3. Why is Health Information(HI) management so hard?
Health
Information Management
Complex health information
Life and death
Confidentiality
Diversity of information
Free Text
Medication handling
Questionnaire, Checklist etc
Images, Videos, Files etc
Diversity among healthcare professionals
Mobile population
Clinical diversity
Continually growing
(HVL) Health Information Exchange November 5, 2019 3 / 34

4. Why do we need to structure HI?
Avoid repetitive data entry
Retrieval and overview
Reuse of record info
For technical integration
For information exchange
Clinical decision support
Quality indicators
Management data
(HVL) Health Information Exchange November 5, 2019 4 / 34

5. How can we structure HI?
By using of consistent standards defining syntactic and semantic meaning
of information being exchanged.
(HVL) Health Information Exchange November 5, 2019 5 / 34

6. Health Information Interchange
REST
API
REpresentational State Transfer
CORBA
Common Object Re-
quest Broker Architecture
SOAP Simple Object Access Protocol
RPC Remote procedure call
(HVL) Health Information Exchange November 5, 2019 6 / 34

7. Issues with RESTful API
Query Complexity
Overfetching
Under-fetching and n+1 request problem
Modifiability
API versioning
(HVL) Health Information Exchange November 5, 2019 7 / 34

8. RESTFul approach - Patients
- Base URL: http://hapi.fhir.org/baseDstu3
GET /Patient All Patients
POST /Patient Create Patient
GET /Patient/649227 Single Patient
PUT /Patient/649227 Update Patient
DELETE /Patient/649227 Delete Patient
https://www.hl7.org/fhir/resourcelist.html lists 143 HL7 FHIR
resources.
Assuming Best Case, each resources needs 5 endpoints (CRUD)
resulting (143 x 5 = 715) endpoints.
Worst Case: Custom endpoints based on custom requirements. Eg.
all patients over 30 years, all patients only from Bergen etc.
(HVL) Health Information Exchange November 5, 2019 8 / 34

9. Self Assessment App
1. (name, description), 2. (name, description, title), 3. Question and options
(HVL) Health Information Exchange November 5, 2019 9 / 34

10. Response From REST - Overfetching
Getting more information than required - overfetching.
Getting less information than required - Underfetching
Need for addition HTTP request to get all the required resource - n+1
request problem
(HVL) Health Information Exchange November 5, 2019 10 / 34

11. Response From REST - Visualization
(HVL) Health Information Exchange November 5, 2019 11 / 34

12. API versioning
the requirements of clients are
often dynamic.
can modify the existing
endpoints or create endpoints
to only fetch required
resources. BUT requirements
changes quickly.
modification quickly becomes
inflexible as one need to think
about how to support existing
customers while providing new
functionality.
(HVL) Health Information Exchange November 5, 2019 12 / 34

13. Our solution
GraphQL
GraphQL developed by Facebook and has been embraced users including
Coursera, GitHub, Pinterest, Neo4J, PayPal and others.
Prototype
Web Client
Authorization
Server
Resource
Server
Mobile Client
(HVL) Health Information Exchange November 5, 2019 13 / 34

14. Architecture Centric Development (ACD)
(HVL) Health Information Exchange November 5, 2019 14 / 34

15. Mapping HL7 FHIR resources to GraphQL schema
Algorithm 1: Mapping HL7 FHIR resources to GraphQL schema
Input: HL7 Resource
Output: GraphQL Schema
1 function recursive_hl7fhir_graphql_mapper (Resource)
2 schema = {};
3 foreach field ∈ HL7 Resource.fields do
4 switch Resource.field do
5 case field.Type is ScalarTypeDefinition do
6 if field.cardinality is 0,1 then
7 add_to_schema(field, type)
8 end if
9 if field.cardinality is 0,* then
10 add_to_schema_as_list(field, type)
11 end if
12 end case
13 case field.Type is EnumTypeDefinition do
14 if EnumTypeDefinition already exists then
15 - reference to schema
16 else
17 - define_new_type_enum(**args)
18 - reference to schema
19 end if
20 end case
21 case field.Type is Custom OR field.Type is HL7 FHIR Resource do
22 if Custom OR Resource already exists then
23 - reference to schema
24 else
25 - define new type Resource
26 - reference to schema
27 - recursive_hl7fhir_graphql_mapper(Resource)
28 end if
29 end case (HVL) Health Information Exchange November 5, 2019 15 / 34

16. Mapping HL7 FHIR with GraphQL schema
Constraint Model
t ype Per son {
name: St r i ng
pi ct ur e( si ze: I nt ) : Ur l
}
Resource
Element
Definition
Data Types
Profile
Conformance
Statement
ValueSet
Code System
Information Model Terminology Model
GraphQL Schema
RootOperationTypeDefinition
Types
ObjectTypeDefinition
ScalarTypeDefinition
EnumerationTypeDefinition
Field
NameDataTypeArguments
<<mapsTo>>
<<mapsTo>>
<<mapsTo>>
schema {
quer y: MyQuer yRoot Type
}
t ype MyQuer yRoot Type {
someFi el d: St r i ng
}
Cardinality
ScalarType EnumType
(HVL) Health Information Exchange November 5, 2019 16 / 34

17. Case Study - Questionnaire
Questionnaire Item AnswerOption
Code
Response
questions answerOptions
responseItems
[0..*][1..*]
Score
[=] answer
[inj]
[1..*]
Class DataType
attribute
reference
hasCode
hasScore
Model, M1
Model, M0
Patient
questionnaire
[1..1]
[1..1]
[1..*]
ResponseItem
source
(HVL) Health Information Exchange November 5, 2019 17 / 34

18. Schema Definition
type Questionnaire {
resourceType: String
url: String
identifier: [Identifier]
version: String
name: String
title: String
status: statusEnumType
experimental: Boolean
publisher: String
description: String
item: [QuestionnaireItem]
}
type Identifier {
system: String
value: String
use: indentifierEnumType
}
type QuestionnaireItem {
linkId: String
prefix: String
text: String
type:
QuestionnaireItemTypeEnum
required: Boolean
answerValueSet: [ValueSet]
}
type ValueSet {
resourceType: String
url: String
identifier: [Identifier]
version: String
name: String
title: String
status: statusEnumType
experimental: Boolean
publisher: String
description: String
compose: Compose
}
type Compose {
inactive: Boolean
include: [ComposeInclude]
}
type ComposeInclude {
concept: [
ComposeIncludeConcept
]
}
type ComposeIncludeConcept {
code: coding
display: String
extension: [Extension]
}
type Extension {
uri: String
valueDecimal: Float
}
enum statusEnumType {
draft
active
retired
unknown
}
enum indentifierEnum {
usual
official
temp
secondary
old
}
enum TypeEnum {
group
display
boolean
decimal
integer
date
dateTime +
}
type Query {
(HVL) Health Information Exchange November 5, 2019 18 / 34

19. Evaluation of Response size and Time
4.07
2.26
2.0
4.0
6.0
8.0
NumberofKilobytes
3.33
1.59
7.42
4.34
4.22
2.5
6.94
3.56
PHQ-9
GAD7
ASRSV1
M
DI
ASRM
S
90.1
79.5
80.0
90.0
100.0
110.0
Timeinmilliseconds
88.1
80.7
109.3
91.9
106.9
85.4
94.5
85.8
PHQ-9
GAD7
ASRSV1
M
DI
ASRM
S
94.5
Kilobytes fetched (REST)
Kilobytes used (GraphQL)
Time to fetch all attributes (REST)
Time to fetch required attributes (GraphQL)
(HVL) Health Information Exchange November 5, 2019 19 / 34

20. Throughput and response time with RESTful approach
(HVL) Health Information Exchange November 5, 2019 20 / 34

21. Throughput and response time with GraphQL
(HVL) Health Information Exchange November 5, 2019 21 / 34

22. Performance Test statistics
Throughput :average number of HTTP/s requests per second
generated by the test
Response Time: average amount of time from first bit sent to the
network card to the last byte received by the client.
(HVL) Health Information Exchange November 5, 2019 22 / 34

23. Challenge: Circular relationship complexity
type ThreadDefinition {
title: String
text(first: Int, after: String): [
MessageDefinition]
}
type MessageDefinition {
text: String
thread: ThreadDefinition
}
type Query {
thread(id: ID!): ThreadDefinition
}
query complicatedQuery {
thread(id: "ID") {
text(first: 100000) {
thread {
text(first: 100000) {
thread {
# ...repeat 100000 (n) times
...
}
}
}
}
}
}
}
(HVL) Health Information Exchange November 5, 2019 23 / 34

24. Challenge: Schema Duplication
schema definition based on the choice of the database being used
(this project uses mongoDB, so schema are based on mongoose 1);
schema definition for a GraphQL endpoint.
Good thing is community is already aware of both issues. And there
are solutions being worked out currently.
1
https://mongoosejs.com/
(HVL) Health Information Exchange November 5, 2019 24 / 34

25. Future Directions
Schema Stitching or Schema federation to promote interoperability
between current health systems and legacy systems.
Creation of adaptive Internet-Delivered Psychological Interventions
(IDPT) using GraphQL based HIE.
Create a comprehensible dashboard for better visualization for
therapists.
Further research in both development of Adaptive system and clinical
trials are required.
(HVL) Health Information Exchange November 5, 2019 25 / 34

26. Conclusion: ISO/IEC 25000.ISO/IEC 25000:2005
Interoperability HL7 FHIR
Security SMART on FHIR
Modifiability SOA-Oriented architecture, GraphQL
Scalability SOA-Oriented architecture
Testability TDD Approach
(HVL) Health Information Exchange November 5, 2019 26 / 34

27. Thank You
skmu@hvl.no
(HVL) Health Information Exchange November 5, 2019 27 / 34

28. HL7 FHIR ER Diagram
Or gani zat i on
i dent i f i er
par t Of
Pat i ent
i dent i f i er
managi ngOr gani zat i on
gener al Pr act i t i oner
Quest i onnai r e
i dent i f i er
answer Val ueSet
Pr act i t i oner
i dent i f i er
Car ePl an
i dent i f i er
subj ect
0. . . *
0. . . 1
1. . . 1
Val ueSet
i dent i f i er
0. . . 1
Quest i onnai r eResponse
basedOn
par t Of
quest i onnai r e
Obser vat i on
i dent i f i er
basedOn
subj ect
0. . . 1
0. . . *
0. . . *
0. . . * 0. . . 1
(HVL) Health Information Exchange November 5, 2019 28 / 34

29. How does GraphQL work?FrontendClient
GraphQLquery
Executiontree
HTTP Request
Query
Mutation
Subscription
BackendOperations
Resolver
Resolver
Resolver
Resolver
Database
DataStore
DBModels
DBConnectors
(HVL) Health Information Exchange November 5, 2019 29 / 34

30. Resolvers
What is a resolver?
A resolver is a function that resolves a value for a type or field in a
schema. If an Object is returned, execution continues to the next child
field. If a scalar is returned (typically at a leaf node), execution completes.
If null is returned, execution halts and does not continue.
Executing queries
Parse: A query is parsed into an abstract syntax tree (or AST).
Validate: The AST is validated against the schema. Checks for
correct query syntax and if the fields exist.
Execute: The runtime walks through the AST, starting from the root
of the tree, invokes resolvers, collects up results, and emits JSON.
(HVL) Health Information Exchange November 5, 2019 30 / 34

31. References I
Duane Bender and Kamran Sartipi.
HL7 FHIR: An agile and RESTful approach to healthcare information exchange.
In Proceedings of CBMS 2013 - 26th IEEE International Symposium on Computer-Based Medical Systems, 2013.
M. Bryant.
Graphql for archival metadata: An overview of the ehri graphql api.
In 2017 IEEE International Conference on Big Data (Big Data), pages 2225–2230, Dec 2017.
ISO/IEC 25000.
ISO/IEC 25000:2005 Software Engineering - Software product Quality Requirements and Evaluation (SQuaRE) - Guide
to SQuaRE, 2005.
R. Srinivasan.
RPC: Remote Procedure Call Protocol Specification Version 2.
RFC 1831, DDN Network Information Center, 1995.
Roy Fielding.
Architectural Styles and the Design of Network-based Software Architectures.
PhD thesis, 2000.
Hl7 FHIR SMART app launch, Retrieved December 28.
Roberto Rodriguez-Echeverria, Javier Luis Cánovas Izquierdo, and Jordi Cabot.
Towards a UML and IFML Mapping to GraphQL.
In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in
Bioinformatics), 2018.
Facebook.
Graphql: A query language for apis, Retrieved April 11, 2019.
(HVL) Health Information Exchange November 5, 2019 31 / 34

32. References II
Ka I Pun Yngve Lamo Suresh Kumar Mukhiya, Fazle Rabbi.
An architectural design forself-reporting e-health systems.
In ICSE 2019 Proceedings in the IEEE Digital Library, 2018.
Hevner, March, Park, and Ram.
Design Science in Information Systems Research.
MIS Quarterly,, 2004.
H. Ulrich, J. Kern, D. Tas, A. K. Kock-Schoppenhauer, F. Ückert, J. Ingenerf, and M. Lablans.
QL 4 MDR: A GraphQL query language for ISO 11179-based metadata repositories.
BMC Medical Informatics and Decision Making, 2019.
Jun Li, Yingfei Xiong, Xuanzhe Liu, and Lu Zhang.
How does web service API evolution affect clients?
In Proceedings - IEEE 20th International Conference on Web Services, ICWS 2013, 2013.
Khalil Khoumbati and Marinos Themistocleous.
Integrating the IT Infrastructures in Healthcare Organisations : a Proposition of Influential Factors.
Electronic Journal of e-Government, 2006.
Yigang Xu, Dominique Sauquet, Eric Zapletal, David Lemaitre, and Patrice Degoulet.
Integration of medical applications: The ’mediator service’ of the SynEx platform.
International Journal of Medical Informatics, 2000.
Olaf Hartig and Jorge Pérez.
Semantics and Complexity of GraphQL .
2018.
(HVL) Health Information Exchange November 5, 2019 32 / 34

33. References III
Nat Sakimura, John Bradley, Michael B. Jones, Breno Medeiros, and Chuck Mortimore.
Final: OpenID Connect Core 1.0 incorporating, 2014.
P. Ferguson and D. Senie.
Network Ingress Filtering: Defeating Denial of Service Attacks which employ IP Source Address Spoofing.
Technical report, 2000.
Cross-origin Resource Sharing.
Cross-Origin Resource Sharing.
Options, 2011.
Suresh Kumar Mukhiya and Khac Hoang Hung.
An Architectural Style for Single Page Scalable Modern Web Application.
5(4):6–13, 2018.
ISO/IEC 25000.
ISO/IEC 25000:2005 Software Engineering - Software product Quality Requirements and Evaluation (SQuaRE) - Guide
to SQuaRE, 2005.
Scott Ambler.
Process Patterns Building Large-Scale Systems Using Object Technology.
1998.
(HVL) Health Information Exchange November 5, 2019 33 / 34

34. A GraphQL approach to Healthcare Information
Exchange with HL7 FHIR
Suresh Kumar Mukhiya, Fazle Rabbi, Ka I Violet Pun,
Adrian Rutle, Yngve Lamo
� skmu@hvl.no
November 5, 2019
(HVL) Health Information Exchange November 5, 2019 34 / 34